AMD is expanding its artificial intelligence (AI) systems as companies move from simply using chatbots to AI that can make decisions, complete tasks, and work with less human help.

At Advancing AI 2026, AMD showed how it is bringing together chips, networking equipment, and software to run these more demanding AI applications.

“The next phase of AI will span frontier models, agents and physical AI, creating new opportunities to bring intelligence everywhere,” said Dr. Lisa Su, chair and CEO, AMD. “Realizing that potential will take the entire industry working together. AMD is partnering across the ecosystem to deliver leadership compute and open platforms that give customers the performance, flexibility and choice to scale AI from the data center to the edge.”

AMD’s main offering is Helios, a large AI computing system designed for companies that need to run AI at scale.

A Helios system packs 72 AMD Instinct MI455X GPUs and 18 6th Gen AMD EPYC “Venice” CPUs. It also uses AMD networking technology and ROCm software.

AMD said Helios can deliver up to 30% more tokens per dollar than the leading competing system. Tokens are the pieces of text AI models process, making the figure a way to compare how much AI work a system can do for the money.

Companies including OpenAI, Anthropic, Meta, Microsoft, Oracle, HUMAIN, Tensorwave, Vultr, and Cirrascale are using or preparing to use AMD’s systems.

Anthropic plans to deploy up to 2 gigawatts of AMD Instinct MI455X GPUs. OpenAI expects to start bringing Helios systems online in the fourth quarter of 2026, with more deployments planned in 2027.

AMD is also making it easier for developers to build AI applications on its hardware.

Its new ROCm.ai platform allows AI coding tools such as Claude, Codex, and Cursor to work directly with AMD’s software and hardware.

AMD is also planning new AI chips through 2030, including Instinct MI500 GPUs in 2027 and MI600 GPUs in 2028.

The company is taking AI beyond large data centers, too. Its Ryzen AI Halo platform is designed for developers who want to run AI locally, while its Kria AI systems are designed for robots that can see, make decisions, and act on their own.

Post navigation